Scaling bug in plotting waveforms/templates for checkerboard-like probes

[zzhmark]

Plotting waveforms for my probe yielding very small waveforms, unless a scale around 1000 is specified. I found the y_scale factor in the plotting functions are very small, due to the layout of my channel positions, which is a non-uniform checkerboard. Some of the y coords may have miniscule differences, and it causes the program to perceive it as a very small inter channel distance. You might need a more adaptive strategy with such scenario.

[zm711]

Absolutely agree. I was having this issue too because of a the nature of the x changes between my contacts on the probe leading to x squashing. I think we need to work on a better solution overall! I think a lot of us are busy. You want to give this a try? Otherwise we will put this on the todo list!

[zzhmark]

Absolutely agree. I was having this issue too because of a the nature of the x changes between my contacts on the probe leading to x squashing. I think we need to work on a better solution overall! I think a lot of us are busy. You want to give this a try? Otherwise we will put this on the todo list!

Here's my plan:

  y_chans, inds = np.unique(channel_locations[:, 1], return_inverse=True)
  if y_chans.size > 1:
      delta_y = 0
      weight = 0
      for i, y_diff in enumerate(np.diff(y_chans)):
          x1 = channel_locations[inds == i, 0]
          x2 = channel_locations[inds == i + 1, 0]
          x_diff = np.min(np.abs(x1[:, np.newaxis] - x2[np.newaxis, :]))
          delta_y += y_diff / (x_diff + 1)
          weight += 1 / (x_diff + 1)
      delta_y = delta_y / weight
  else:
      delta_y = 40.0

Considering very close y contacts, I'm introducing a weighting strategy punishing contacts close over y but far over x.

If you think this is ok I will submit a merge request.

[zm711]

@alejoe91 and @samuelgarcia review the widgets stuff so I'll tag them to take a look at your proposal.